Researcher

Oscar Beijbom

Oscar Beijbom is a computer vision and machine-learning researcher whose work applies deep learning to coral monitoring and autonomous driving. He earned an engineering-physics M.S. at Lund and a computer-vision Ph.D. at UCSD, then developed coral-image systems at Berkeley and coauthored nuScenes and nuPlan. He directed machine-learning work at Motional and co-founded Nyckel, an API company for custom machine learning. As of October 2026, he is a Principal Engineer at Zoox, working on perception, planning and foundational modeling for autonomous vehicles.

Now at
Zoox

Timeline

  1. 2007–2015

    Lund and UCSD degrees

    He completes a master's thesis on single-image focus assessment with support vector machines at Lund in 2007 and earns a Ph.D. in computer vision and machine learning from UCSD in 2015 under David Kriegman and Serge Belongie. His M.S. in engineering physics at Lund is under Kalle Åström.

  2. 2014

    Cost-sensitive multiclass boosting

    In 2014, he coauthors an ICML paper on cost-sensitive multiclass boosting.

  3. 2016–2017

    Deep learning for coral reefs

    In 2016, he works as a postdoctoral scholar in Trevor Darrell's Berkeley Vision and Learning Center, where his neural-network research helps coral researchers map reefs. In 2017, CoralNet's Beta software uses a deep neural network for reef-image annotation.

  4. 2018–2019

    nuScenes for autonomous driving

    He coauthors nuScenes, a multimodal autonomous-driving dataset with six cameras, five radars and one lidar providing full 360-degree coverage. Its teaser set is released in September 2018, and the full dataset is released in March 2019 with 1,000 annotated 20-second scenes.

  5. 2020

    Director at Motional

    In 2020, he is Director of Machine Learning at Motional, where he works on products, research, data curation and the ML platform. Earlier, he leads development at Hovding and creates the algorithmic framework and hardware design for its invisible bicycle helmet.

  6. 2021

    CoralNet and nuPlan

    In 2021, he coauthors a new CoralNet deep-learning engine with EfficientNet-B0; the team reports 22.4% lower average error than CoralNet Beta on a 43,049-image test set. He also coauthors nuPlan, an ML-based closed-loop planning benchmark for autonomous driving.

  7. 2023

    Co-founding Nyckel

    In 2023, he is co-founder and CTO of Nyckel, which builds custom machine-learning APIs for non-expert users. Its service trains and evaluates models on uploaded data and returns predictions.

  8. As of October 2026

    Principal Engineer at Zoox

    As of October 2026, he is a Principal Engineer at Zoox, where he works across perception, planning and foundational modeling for autonomous vehicles.

Career

  1. NowZooxPrincipal Engineer
  2. From 2023NyckelCo-founder, CTO
  3. From 2020MotionalDirector of Machine Learning
  4. 2016–2017University of California, BerkeleyPostdoctoral Scholar

Education

  1. 2015University of California, San DiegoPh.D. in Computer Vision and Machine Learning
  2. 2007Lund UniversityM.S. in Engineering Physics

Sources and credits

Photo: We have no published photo of Oscar Beijbom yet. Send us a free-licensed photo.

  1. CS 201: Data, Detection, and Metrics for Autonomous Vehicles, OSCAR BEIJBOM, Motional cs.ucla.edu. Checked 20 October 2026.
  2. Oscar Beijbom | Vision vision.ucsd.edu. Checked 20 October 2026.
  3. Software system labels coral reef images in record time cse.ucsd.edu. Checked 20 October 2026.
  4. Research Archives - Page 88 of 91 - EECS at Berkeley eecs.berkeley.edu. Checked 20 October 2026.
  5. nuScenes: A multimodal dataset for autonomous driving arxiv.org. Checked 20 October 2026.
  6. A new deep learning engine for CoralNet coralnet.ucsd.edu. Checked 20 October 2026.
  7. NuPlan: A closed-loop ML-based planning benchmark for autonomous vehicles arxiv.org. Checked 20 October 2026.
  8. Oscar Beijbom, co-founder and CTO of Nyckel, on the opportunites in the AI/ML tooling market sacra.com. Checked 20 October 2026.
  9. Oscar Beijbom beijbom.github.io. Checked 20 October 2026.